Design, development, and validation of a high-throughput drug-screening assay for targeting of human leukemia




The authors developed an ex vivo methodology to perform drug library screening against human leukemia.


The strategy for this screening relied on human blood or bone marrow cultures under hypoxia; under these conditions, leukemia cells deplete oxygen faster than normal cells, causing a hemoglobin oxygenation shift. Several advantages were observed: 1) partial recapitulation of the leukemia microenvironment, 2) use of native hemoglobin oxygenation as a real-time sensor/reporter, 3) cost-effectiveness, 4) species specificity, and 5) a format that enables high-throughput screening.


For a proof of concept, a chemical library (size, approximately 20,000 compounds) was screened against human leukemia cells. Seventy compounds were identified (“hit” rate, 0.35%; Z-factor = 0.71) that had activity, and 20 compounds were examined to identify 18 true-positive compounds (90%). Finally, the results demonstrated that carbonohydraxonic diamide group-containing compounds are potent antileukemia agents that induce cell death in leukemia cells and in patient-derived samples.


The current results indicated that this unique functional assay can identify novel drug candidates and can help with the development of future applications in personalized drug selection for patients with leukemia. Cancer 2014;120:589–602. © 2013 American Cancer Society.


Functional screening platforms that can assess drug candidates within the appropriate tumor microenvironment are needed. First, mouse bone marrow is generally less sensitive to many cytotoxics than human bone marrow, thus, often rendering murine leukemia models inadequate for emulating myelosuppression in patients.[1] Second, the coexistence of nonmalignant host cells (immune, stromal, mesenchymal) and leukemia stem cells provokes complex tumor-host interactions that may affect drug resistance, self-tolerance, angiogenesis, tumor growth, and response to therapy.[2] Species-specific differences between human leukemia cells and mouse stromal cells can alter tumor growth and drug responses in murine models,[3] and deficiency of the functional immune system in these models may also interfere with the outcome.[2]

Ex vivo models using human tissue as a screening platform are valuable preclinical tools. In human solid tumors, multicellular tumor spheroid models that were able to recapitulate in vivo-like growth have proven to be excellent in vitro 3-dimensional models for high-throughput drug discovery.[4] Here, we describe a new, functional, high-throughput, ex vivo screening assay against leukemia that is based on culturing leukemia cells in human blood or bone marrow under hypoxic conditions. We reasoned that these cocultures mimic the disease microenvironment and, thus, partially recapitulate at least some attributes of leukemia in patients. Moreover, the oxygenation state of native hemoglobin reliably and reproducibly serves as a “built-in” indicator of leukemia cell growth and/or viability, therefore overcoming the need for elaborate detection methods in a multicellular setting.

For a proof of concept, we have used this assay for a chemical library screening on established leukemia cell lines to select “microenvironment-stable” drugs with potential for translation into clinical applications. By using this assay, we identified a subset of carbonohydraxonic diamide group-containing compounds that markedly and specifically inhibited several leukemia cell lines and a panel of clinical samples obtained from leukemia patients. Together, these data suggest that testing of libraries of compounds or candidate drugs in this new ex vivo model may yield compounds against human leukemia that are potentially active in the circulation and/or bone marrow microenvironment.


Cell Culture

OCI-AML3, Kasumi-1, THP-1, HL-60, MOLT-4, CCRF-CEM, HL-60, RPMI-8226, SR-786, U937, KBM7, K562, and K562-luc2 Bioware Ultra cell lines (Caliper LifeSciences, Hopkinton, Mass) were maintained in humidified hypoxia chambers (HeraCell 150; Thermo Electron Corporation/Thermo Fisher Scientific, Waltham, Mass) with 5% CO2 and 5% oxygen at 37°C in RPMI-1640 medium (RPMI) containing 10% fetal bovine serum, penicillin, and streptomycin.

Blood and Bone Marrow Samples From Patients With Leukemia and Normal Volunteers

The Institutional Review Board of The University of Texas MD Anderson Cancer Center approved the use of whole blood and bone marrow samples obtained from patients or healthy donors. Peripheral blood and bone marrow samples were obtained from patients with acute myeloid leukemia (AML) who had signed an informed consent form in accordance with the Declaration of Helsinki. Blood samples from healthy volunteers were obtained through the hospital's Blood Bank and Transfusion Services. We used anonymized blood samples, which had been tested previously and proven negative against blood-transmittable diseases. These samples were stored at 4°C for 24 hours before use. Heparin was used as an anticoagulant. Pretested whole blood and bone marrow samples also were obtained from commercial sources (Innovative Research [Novi, Mich] or AllCells [Alameda, Calif]).

Assays Containing Human Peripheral Whole Blood and Bone Marrow

Leukemia cells were plated at 20,000 per well in 100 μL of RPMI containing 10% human whole blood, heparin (100 μg/mL), L-glutamine (0.292 mg/mL), penicillin (100 U/mL), and streptomycin (100 U/mL) in 96-well plates with flat-bottomed wells (Becton Dickinson and Company, East Rutherford, NJ). Ten-percent blood specimens from patients with AML and 5% to 10% bone marrow aspirates were diluted in either RPMI or complete StemPro TM-34SFM (GibcoBRL/Life Technologies, Grand Island, NY) culture medium containing heparin, L-glutamine, penicillin, and streptomycin. The microplates were incubated under hypoxia (without shaking) and the optical density at 600 nm (OD600) was measured at the starting point and after 20 hours and/or 40 hours of incubation. The Micros60 analyzer (ABX Diagnostics Inc., Irvine, Calif) was used to count white blood cells (WBCs), granulocytes, monocytes, lymphocytes, erythrocytes (red blood cells [RBCs]), platelets, hemoglobin, hematocrit, mean corpuscular volume, mean corpuscular hemoglobin, mean corpuscular hemoglobin concentration, and RBC distribution width.

Chemical Library and Drug Screening Against Leukemia Cell Lines

We used the DiverSet Chemical Library (ChemBridge Corporation, San Diego, Calif) formatted in 96-well plates and containing small-molecule compounds with drug-like properties. We screened 20,000 individual compounds from the chemical library, each at 20 μM, against OCI-AML3 cell lines. Primary compounds that decreased the OD600 by at least 0.2 units were selected for secondary screening and analysis; moreover, structural analogues with at least 50% similarity to the primary compounds were commercially obtained (ChemBridge Corporation) and subsequently evaluated.

Leukemia Proliferation, Viability, and Cell Death Assays

Proliferation of luciferase-transfected K562 leukemia cells was determined with substrate D-luciferin (Xenogen Biosciences, Alameda, Calif) incubated at 150 μg/mL per well for 1 hour followed by measurement of the luminescence (SpectraMax 5 or SoftMax Pro 5; Molecular Devices LLC, Sunnyvale, Calif). Cell proliferation and viability were measured with a lactate dehydrogenase (LDH) activity assay (DHL; AnaSpec Inc., Fremont, Calif). To measure the incorporation of 5-bromo-2′-deoxyuridine (BrdU) (Calbiochem/EMD Millipore, Billerica, Mass), cells were incubated for 2 hours with BrdU, RBCs were lysed with lysis buffer (Roche Diagnostics Inc., Indianapolis, Ind), and the remaining cells were pelleted and fixed for immunodetermination of incorporated BrdU. For the assessment of cell apoptosis and necrosis, in total, 106 OCI-AML3 cells were incubated for 4 hours at 37°C in hypoxia status in a volume of 5 mL in the presence of each test compound at 20 μM. After washing, the cells were stained with a fluorescein isothiocyanate (FITC)-conjugated annexin V antibody plus propidium iodide (Sigma Chemical Company, St. Louis, Mo) and subsequently analyzed by flow cytometry (BD Canto II; BD Biosciences, San Jose, Calif).

Statistical Analysis

All data are reported as mean ± standard deviation (SD) values. Student t tests (unpaired) were used to determine statistical significance (n = 3, unless otherwise specified). P values that were considered statistically significant are indicated in the figures by asterisks (a single asterisk indicates P < .05; double asterisks, P < .01; and triple asterisks, P < .001). Correlations were calculated with the Pearson correlation coefficient. The Z-factor was calculated according to the described formula.[5]


Real-Time Leukemia Cell Monitoring in an Ex Vivo Blood-Containing Assay

Cells from a panel of leukemia cell lines (K562, OCI-AML3, Molt-4, THP-1, and Kasumi-1 cells) were able to proliferate in medium that contained 1:10 (volume/volume) whole blood from healthy donors under 5% oxygen (data not shown). Under these conditions, live leukemia cells induced a hemoglobin color switch to dark purple. The hemoglobin purple color switch was reversible to bright red when the samples were returned to a 21% oxygen condition, and shaking accelerated this change, a result indicating that the color change depended on the oxygenation state of hemoglobin (Fig. 1A). We hypothesized that this assay may be a suitable platform for screening of antileukemia compounds that are stable in the presence of blood. A schematic presentation of the screening is provided in Figure 1B along with an example of a 96-microwell plate containing a newly discovered leukemia cell inhibitor (Fig. 1C).

Figure 1.

The monitoring of leukemia cell viability in a human blood-containing assay is illustrated. (A) This is a schematic illustration of oxygen consumption of leukemic cells in the presence of blood. The oxygen consumption of metabolically active leukemia cells is observed as a color change in the presence of blood under hypoxia, as leukemia cells deplete the surrounding oxygen, causing hemoglobin to shift to its deoxyhemoglobin conformation. RBC indicates red blood cells (erythrocytes). (B) This schematic shows a large chemical library screening, with bright red wells displaying leukemia cell cultures that were not metabolically active. (C) A 96-well plate is shown with different compounds tested in each well. The arrow indicates a compound of interest, which inhibited the color shift and, thus, was selected for further studies.

Optimization of the Blood-Containing Ex Vivo Assay for Leukemia Cells

Time-dependent proliferation of OCI-AML3 cells in the ex vivo cultures was demonstrated with a blood count analyzer, which demonstrated an increase in total WBCs (Fig. 2A). In addition to the hemoglobin-mediated color switch, leukemia cell status also was evaluated by 3 other independent assays: cellular LDH activity, BrdU incorporation, and activity of transfected luciferase (Fig. 2B-D). Leukemia cell growth above background became detectable after culturing for 20 hours or longer. We used whole blood samples from more than 200 individual healthy donors, all of which supported leukemia cell growth.

Figure 2.

Charts illustrate leukemia cell cultures in the presence of human whole blood or bone marrow aspirate. The K562 or OCI-AML3 cell lines were cultured at 20,000 cells per well in a culture medium containing 10% normal human blood or bone marrow aspirate, as described in the text. (A) An increase in leukemia cell counts in the presence of blood was determined by measuring total white blood cell (WBC) counts. (B) Cell viability and growth were determined by measuring lactate dehydrogenase (LDH) activity, and (C) dividing cells were assessed using bromodeoxyuridine (BrdU) incorporation (+ indicates positive; −, negative). (D) Increase in luciferase activity was measured in stably transfected K562-luc2 cells. (E) The color shift was detected according to the OD600 value by measuring it in both hypoxia and normoxia cultures grown for 40 hours. To assess the dependence of the color shift on the oxygen tension, both cultures were oxygenated (by gently pipetting up and down), and the OD600 value was measured again. (F) LDH activity was compared between cells grown in hypoxia and normoxia for 40 hours. (G) Increases in leukemia cell counts in the presence of bone marrow aspirate were determined by measuring total WBC counts. (G) The OD600 of cultures with bone marrow aspirates was determined. (I) The absorbances were read again after oxygenating the cultures as previously described. (J) Etoposide (10 μM), which inhibits the color shift because of its antileukemia activity, served as a positive control. The results represent mean ± standard deviation values from triplicate wells. A single asterisk indicates P < .05; double asterisks, P < .01; triple asterisks, P < .001.

OD600 measurements were substantially higher in leukemia ex vivo assays cultured in hypoxia rather than in normoxia after 40 hours (Fig. 2E). Next, both cultures were oxygenated by mixing, and the differences in OD600 measurements virtually disappeared, indicating that a higher OD600 can be observed after the leukemia cells deplete oxygen from the microenvironment. Background levels of OD600 varied little (range, 1.7-1.9 units) at 0 hours. The OD600 increased less than 0.2 units in leukemia-free control assays after incubation for 40 hours; in contrast, the presence of leukemia cells generally increased the OD600 by at least 0.4 to 0.6 units (Fig. 2E). To exclude the possibility that different cell growth rates in hypoxia versus normoxia affected the OD600 measurements, we demonstrated that leukemia cells grew similarly in both conditions (Fig. 2F). In addition to normal peripheral blood, we also established that leukemia cells could be cultured in medium containing human bone marrow samples obtained from healthy volunteer donors (Fig. 2G). In those samples, the oxygen consumption of live cells also could be determined by measuring the OD600 (Fig. 2H). Correspondingly, the OD600 decreased upon oxidation of these cultures (Fig. 2I). To confirm that the blood-containing assay could be used to detect antileukemia compounds, we used etoposide, a topoisomerase II inhibitor with known cytotoxic activity,[6] as a positive control (Fig. 2J).

To verify the dependence of OD600 values on leukemia cell viability and growth, we determined both OD600 values and WBC counts in 2 different donor blood samples in the presence or absence of several antileukemia compounds. The increase in OD600 values, as measured by the blood count analyzer, was correlated with the increase in total WBC counts (r = 0.81 and r = 0.79 for donors 1 and 2, respectively); granulocytes (r = 0.82 and r = 0.84, respectively); and, to a lesser extent, monocytes (r = 0.66 and r = 0.65, respectively) (Fig. 3). Therefore, we concluded that the absorbance of hemoglobin[7, 8] at the OD600 may be reliably used as an initial indicator of leukemia viability and/or cell growth in these assays.

Figure 3.

Charts illustrate the correlation between the optical density at 600 nm (OD600) and counts determined by automatic blood analyzer of total white blood cells (WBCs) and the subsets of granulocytes and monocytes in 2 different donor blood specimens. Correlation was determined by culturing OCI-AML3 cells in 10% blood for 20 hours and 40 hours in the absence and presence of inhibitors (n = 10 for donor 1, n = 7 for donor 2; each at 10 μM and 20 μM) that had been discovered previously in the screening process.

Screening of Antileukemia Compounds in the Presence of Human Whole Blood

Next, as a proof of concept, we screened a 20,000 small-molecular-weight compound library (ChemBridge Corporation) against OCI-AML3 leukemia cells. We observed an approximately 0.35% “hit rate” to identify 70 potential antileukemia compounds (data not shown) that decreased the OD600 by at least 0.2 units. The Z-factor,[5] which was calculated for positive hits against background in a particular blood sample, was 0.71 (n = 30). We examined a subset of 20 randomly selected compounds out of the initial “pool” and demonstrated that 18 of 20 compounds (90%) retained activity against OCI-AML3 cells at 10 μM concentration in standard culture conditions in the absence of blood (data not shown). Next, we selected 4 structurally different test compounds out of the 18 active candidates for subsequent studies (termed compounds #1, #2, #3, and #4), and compared their efficacy to inhibit leukemia cells in the presence of blood by using different cell proliferation/viability detection methods (Fig. 4A). Of these, the compound #1 (N″-[4-([4-bromo-2,3,5,6-tetramethylbenzyl]oxy)-3-methoxy-benzylidene] carbonohydrazonic diamide hydrochloride) had the most robust activity. To identify the chemical group(s) required for the antileukemia activity, we evaluated a panel of structural analogues of compounds #1, #2, #3, and #4 with the criterion that similarity to the parent compound must be at least 50%. These compound analogues were tested on both the OCI-AML3 and K562 leukemia cell lines in the blood-containing assay (Fig. 4B) and on other leukemia and lymphoma cell lines in standard culture (Fig. 5). These experiments demonstrated that 5 analogues of compound #1, all of which contained a carbonohydrazonic diamide group, were the most effective inhibitors of leukemia cells in both the presence and the absence of blood. Several analogues that lacked carbonohydrazonic diamide or that had other modifications were inactive in the presence of blood. The statistically significant result (Fisher exact test; P = .0217) indicated that carbonohydrazonic diamide is required for the growth inhibitory activity of this class of compounds.

Figure 4.

Charts and a heat map illustrate evaluation of the 4 selected primary compounds and their analogs. (A) Activities of the compounds were compared with the optical density at 600 nm (OD600), lactate dehydrogenase (LDH) activity, bromodeoxyuridine (BrdU) incorporation, and luciferase assays. OCI-AML3 or K562 cells were cultured in RPMI containing 10% human blood. A single asterisk indicates P < .05; double asterisks, P < .01; triple asterisks, P < .001. DMSO indicates dimethyl sulfoxide. (B) This is a heat-map presentation of the activities of the 4 selected primary compounds and their corresponding structural analogs in OCI-AML3 or K562 cells cultured in 10% blood. Two different donor bloods were used. For compound activity, − indicates no activity; +, Δ OD600≤0.1; ++, Δ OD600≤0.2; +++, Δ OD600≤0.3.

Figure 5.

Charts illustrate activities of the selected structural analogs on a representative panel of human leukemia and lymphoma cell lines (n = 10) in standard culture in the absence of blood. Compounds are numbered according to Figure 4B. Cell viability and proliferation were determined using a lactate dehydrogenase LDH assay. The results represent mean ± standard deviation values from triplicate wells. AML indicates acute myeloid leukemia; CML, chronic myeloid leukemia; T-ALL, T-cell acute lymphoblastic leukemia.

Effects of Carbonohydrazonic Diamide-Containing Compounds on Leukemia Cells

We further analyzed a panel (n = 7) of carbonohydrazonic diamide-containing compounds along with the negative control compound #1N (Fig. 6A). We observed that the carbonohydrazonic diamide-containing compounds inhibited an increase in the OD600 and in leukemia cell counts and reduced the WBC count to background levels at the 20-hour time point with from <0.05% to 0.001% probability (Fig. 6B). Next, we examined the induction of apoptosis in OCI-AML3 cells by these compounds. The analog compounds #1A, #1C, #1D, and #1E induced apoptosis more efficiently (from >70% to 80%) than the original compound #1 (>50%), as determined by annexin V and propidium iodide staining after a 4-hour incubation (Fig. 6C). The compound #1B produced only a slight increase (approximately 5%) in apoptosis compared with control cells. The control compound #1N, which did not contain carbonohydrazonic diamide, did not induce apoptosis.

Figure 6.

Carbonohydraxonic diamide-containing compounds are illustrated. (A) Chemical structures of 6 active carbonohydraxonic diamide-containing compounds are shown. The carbonohydrazonic diamide group is depicted in blue, and differences in structures between different compounds are depicted in red. (B) The effects of carbonohydrazonic diamide-containing compounds (at 10 μM each) are plotted for OCI-AML3 cells that were cultured in media containing human blood. The optical density at 600 nm (OD600) and white blood cell (WBC) counts were determined at the 20-hour time point. The results represent mean ± standard deviation values from triplicate wells. (C) Induction of cell apoptosis and necrosis by carbonohydrazonic diamide-containing compounds is indicated. Cells were analyzed by flow cytometry after staining with annexin V-fluorescein isothiocyanate (FITC) and propidium iodide. DMSO indicates dimethyl sulfoxide. A single asterisk indicates P < .05; double asterisks, P < .01; triple asterisks, P < .001.

Next, we analyzed the effects of carbonohydrazonic diamide-containing compounds on different blood cell populations at 0 hours, 20 hours, and 40 hours using the blood cell counter. In general, OCI-AML3 leukemia cell growth increased the total WBC count by approximately 2-fold after incubation for 40 hours, and the increase was detected mainly in the granulocyte population and, to a lesser extent, in the monocytes. The assay analysis of compound #1A is provided as an example: the compound #1A at 10 μM prevented the increase in WBC counts and had no clear inhibitory effect on lymphocytes or RBCs (Fig. 7A).

Figure 7.

Charts illustrate the effects of compound £1A on individual blood cell populations. (A) OCI-AML3 cells were cultured at 20,000 cells per well in culture medium containing 10% blood in the absence or presence of compound £1A (10 μM). White blood cells (WBCs), lymphocytes, monocytes, granulocytes, and erythrocytes (red blood cells [RBCs]) were determined using a blood cell counter at the time points 0 hours, 20 hours, and 40 hours. OD600 indicates the optical density at 600 nm. (B) OCI-AML3 cells were cultured at 20,000 cells per well in culture medium containing normal human bone marrow aspirate (5% or 10%) in the absence or presence of compound £1A (10 μM or 20 μM). Bone marrow without treatment served as a negative control. The OD600 value was determined at the time points 0 hours, 20 hours, and 40 hours. WBCs, RBCs, and platelets were determined. The results represent mean ± standard deviation values from at least triplicate wells.

Carbonohydrazonic diamide-containing compounds also inhibited leukemia cells in the assay with aspirated human bone marrow. Like in the blood-containing assay, the example compound #1A (Fig. 7B) inhibited an increase in the OD600 and in leukemia cell counts in the bone marrow-containing assay. It is noteworthy that the compound #1A (at 10 μM) sufficed to inhibit leukemia cell growth in the blood-containing medium, but 20 μM were required for efficient inhibition in the bone marrow-containing medium, even when adjusted for the lower volume/volume used (ie, 5% for bone marrow rather than 10% for blood).

Carbonohydrazonic Diamide-Containing Compounds Inhibit Primary Acute Myeloid Leukemia Cells

To validate the translational potential of the blood assay, we used peripheral blood obtained from patients with AML and observed that carbonohydrazonic diamide-containing compounds inhibited primary leukemia cell growth and/or viability in the assays with patient blood. Specifically, of the 6 compounds that had activity against leukemia cell lines, compounds #1A and #1B, which differed only by 1 methyl group, proved to be the most effective, reducing the OD600 in all samples evaluated (Fig. 8A). Notably, the activities of compounds #1, #1C, and #1E varied in different AML samples, suggesting that the efficacy of these compounds may be different in every patient. In contrast, the compound #1D was without detectable effects against primary AML. The negative control #1N did not reduce the OD600 (Fig. 8A), whereas the positive controls (etoposide, actinomycin D, and staurosporin) consistently reduced the OD600 significantly (Fig. 8B). To determine whether the effects of the carbonohydrazonic diamide-containing compounds were leukemia-specific, we measured their effects in ex vivo assays of normal blood from healthy donors. In 2 of 3 normal blood samples, the compounds had no statistically significant effects; however, in the third sample, the compound #1A reduced the OD600 slightly, suggesting individual-specific differences in susceptibility (Fig. 8C).

Figure 8.

Charts illustrate the inhibition of primary acute myeloid leukemia (AML) cells by carbonohydraxonic diamide-containing compounds. (A) Blood samples from patients with AML were diluted to 10% and incubated in the absence or presence of compounds, as indicated (at 10 μM), for 20 hours, and the optical density at 600 nm (OD600) was measured. The initial number of blasts (blast £) (103/μL), blast percentage (blast %), white blood cell count (WBC) (103/μL), and erythrocyte count (RBC) (106/μL) are indicated for each patient sample. (B) Etoposide, actinomycin D, and staurosporin (at 10 μM), which have cytotoxic activities, served as positive controls for the patient blood assay. (C) The effect of carbonohydrazonic diamide-containing compounds on OD600 values in blood from healthy donors was measured after 20 hours of cultivation. (D) The effects of blast number and sample volume on OD600 values are illustrated. Mononuclear cells were isolated and reimplanted into the patient blood samples. The initial blast number (blast £) (103/μL), blast %, WBC count (103/μL), and RBC count (106/μL) are indicated for both patient samples. (E) The efficacy of carbonohydrazonic diamide-containing compounds on primary AML cells under standard tissue culture conditions is illustrated. Cells were cultured for 24 hours; then, viability was determined using a lactate dehydrogenase assay. The results represent mean ± standard deviation values from at least triplicate wells. A single asterisk indicates P < .05; double asterisks, P < .01; triple asterisks, P < .001.

To confirm that the increase in OD600 also correlated with the WBC count in the patient sample setting, we tested the effect of leukemia cell counts on the OD600. WBCs from AML patient samples were isolated and then reinserted into the ex vivo cultures at concentrations that enriched the WBC count by approximately 2-fold. This resulted in a corresponding increase in the OD600, indicating that the WBC counts are a fundamental factor determining the OD600 (Fig. 8D).

Finally, the compounds were evaluated on primary AML cells cultured in standard culture media. In these conditions, all carbonohydrazonic diamide-containing compounds, except #1D, were efficient inhibitors of primary cells isolated from patients with AML (n = 3) (Fig. 8E). In effect, isolated primary AML cells grown in the absence of blood yielded results similar to those obtained with established leukemia and lymphoma cell lines in standard culture conditions (Fig. 5).


In this study, we introduce a novel ex vivo assay for leukemia that provides a high-throughput screening platform for the identification of compounds that are active in the presence of human blood and bone marrow under hypoxia. This assay allows the detection of live leukemia cells by using the hemoglobin oxygenation state as an internal readout system that can be measured using the OD600 based on hemoglobin absorption spectra[7, 8] and, thus, does not require the introduction of either chemically active exogenous markers to measure leukemia cell growth or oxygen probes to estimate cancer cell metabolic rates. Specifically, we observed that viable human leukemia cells would rapidly deplete oxygen from the medium and predominantly contribute to the high OD600 values detected; in contrast, control normal blood incubated with no leukemia cells yielded only slightly elevated levels. This increase in OD600 values correlates well with the increase in leukemia cell counts, indicating that the levels of deoxyhemoglobin in the medium increase in accordance with the oxygen consumption by leukemia cells. Moreover, the OD600 values could be manipulated with the oxygen levels—independent of the other variables—confirming that the oxygenation state of the hemoglobin directly contributed to the OD600 value measured.

Empirically, we detected clear differences in the drug activities of several individual compounds when leukemia cells were grown in standard cultured media versus the blood-containing medium. Thus, we focused this original study on the discovery of drugs with robust activity in the presence of human whole blood. The “Z-factor”—a coefficient that reflects both the assay signal dynamic range and the data variation associated with the signal measurements[5]—was calculated at approximately 0.7 for our assay, which indicates that the results are reproducible and accurate. It is noteworthy that differences among donor blood samples can contribute to observed variations in efficacy of drug candidate evaluations.

For our assay, as an initial proof of concept, we screened a library of approximately 20,000 individual chemical compounds against the OCI-AML3 cell line. We identified 70 lead compounds that had activity against the human leukemia cell lines (library/assay “hit rate,” 0.35%); most of those compounds had not been described previously in the literature, but a few of them had structural similarity to compounds with anticancer, anti-inflammatory, or antimicrobial activity. A few specific examples of promising drug leads merit mention. Compound #2 is active against the blood-borne protozoan parasite Trypanosoma cruzi,[9] and another compound selectively induces apoptosis in tumor cells independent of P-glycoprotein status.[10] We note that certain structural moieties were common within the discovered compounds, such as piperazine, hydrazide, and hydrazone groups. There also have been previous reports of antitumor activity by some of these derivatives.[11] Moreover, from the 70 “hits” in our initial screen, we re-examined an arbitrarily chosen subset of 20 compounds and observed that 18 of those compounds (90%) were reproducibly true-positive for antileukemia activity.

Several limitations of this proof of concept for assay development and candidate drug discovery study merit further comment. First, although specificity and toxicity of the selected drugs are beyond the scope of this initial work, we observed the requirement of relatively high molar concentrations; this may have been caused, at least in part, by a loss of drug activities in the chemical library setting secondary to the presence of impurities and salts or degradation (ie, in materials not of Good Manufacturing Practices grade). Indeed, the simplest carbonohydraxonic diamide, aminoguanidine, has already been examined in clinical trials as a potential antidiabetic agent,[12] suggesting that such compounds may be tolerated by mammals in vivo. To begin addressing this possibility, we have de novo resynthesized compound #1A and observed that it is active and stable at low micromolar concentrations in the presence of whole blood, suggesting that translational applications with this compound actually may be feasible. Moreover, our whole blood assay is certainly not sufficient to address the issue of cell specificity (eg, normal vs malignant cells or leukemia vs nonleukemia tumor cells); comprehensive toxicology evaluation in bone marrow, lymphatic vasculature, as well as other nonhematopoetic organs will be needed to assess the safety and potential therapeutic applications of any of these compounds. Finally, whether or not the growth-inhibitory effect of carbonohydraxonic diamide group-containing compounds is specific remains an open question. Further mechanistic studies will be needed to determine the specificity of carbonohydraxonic diamide-containing compounds toward AML cells, because several potential molecular targets for these types of compounds have been reported, such as glucose-mediated protein dimerization,[12] nitric oxide synthase,[13] furin,[14] and E2 ubiquitin-conjugating enzyme.[15, 16] Ultimately, formal toxicology Good Laboratory Practice studies in animals will be required in the future to determine whether or not a first-in-human and/or phase zero clinical trial will go forward.

Our results indicate that the unique functional drug-screening assay introduced here has the capability to identify novel microenvironment-stable drug candidates. This assay may also help with the development of future applications in personalized drug selection for patients with leukemia as therapeutic compound responses are correlated with specific cancer genotypes,[17] which can greatly vary from individual to individual. In effect, in the pilot experiments, we were able to optimize the blood-containing assay for blood samples directly from patients with AML. Furthermore, similar studies with human bone marrow cocultures demonstrated promise that drug efficacy also could be validated individually in a bone marrow-like microenvironment, in which resistance is often encountered, possibly because of increased protection of tumor cells by bone marrow stroma.[2] Together, our results support the idea that this new methodology could potentially be used as a tool in predicting drug efficacy and/or response in each leukemia patient.


This work was supported by the Specialized Program in Research Excellence (SPORE) in Leukemia from The University of Texas MD Anderson Cancer Center (R.P. and W.A), the Gillson-Longenbaugh Foundation (R.P. and W.A), and the Cancer Society of Finland (E.K.).


Dr. Karjalainen is supported by a Specialized Program in Research Excellence (SPORE) in Leukemia grant as well as grants from the Gillson-Longenbaugh Foundation (money from Renata Pasqualini and Wadih Arap) and the Cancer Society of Finland (money from Erkki Koivunen). She is also an employee of The University of Texas MD Anderson Cancer Center. Drs. Pasqualini and Arap are founders and paid consultants to AAVP Biosystems Inc., Ablaris Therapeutics Inc., Alvos Therapeutics Inc., AMP Pharmaceuticals, and Ceramide Therapeutics LLC and own stock and hold equity in those companies; and they receive license royalty payments from APAvadis Biotechnologies, Merck & Company, and n3D Biosciences Inc. Dr. Cortes receives compensation as a board member of Tragara Pharmaceuticals Inc. and receives consulting fees from Ariad Pharmaceuticals Inc., Pfizer Inc., and Tera Pharmaceutical Industries Ltd. Dr. Sidman has received consulting fees from Drs. Pasqualini and Arap at The University of Texas MD Anderson Cancer Center for writing and editing articles, has received compensation from Sanofi-Genzyme Corporation for consulting on writing and laboratory research, and has received grants from the National Institutes of Health. Dr. Kantarjian has brought in research funding to The University of Texas MD Anderson Cancer Center from Bristol-Myers Squibb, Pfizer Inc., Ariad Pharmaceuticals Inc., and Novartis Pharmaceuticals Corporation.